What was changed
Chinese AI labs faced a basic tradeoff: open-source models could spread faster, but releasing the model for free made direct revenue harder. Moonshot AI chose to release Kimi’s model openly instead of keeping every capability behind a commercial interface.
The decision did not make the whole business free. Rest of World reported that Kimi’s API and subscription remained in high demand because Moonshot had the infrastructure to run the model well. The open release therefore worked as distribution and branding, while the operating layer stayed monetizable.
What it achieved
Moonshot kept API and subscription demand while Kimi’s free model expanded the lab’s reach and gave it a day-zero support advantage.
Why it worked
Open weights let developers download, inspect, and customize the model without depending on a single closed provider.
Kimi’s release built public proof of the lab’s technical ability and helped attract attention and talent.
Users still needed engineering and reliable infrastructure to run an open model, leaving Moonshot a paid service advantage.
What can be applied
Give away the layer that builds adoption and reputation, but keep scarce infrastructure or premium capability as the layer that pays for continued development.
Aftermath
Moonshot’s choice illustrated a hybrid path for AI labs: openness can build adoption and brand, while APIs, subscriptions, infrastructure support, or a stronger base model can fund the work. The same economics also leave open-source labs exposed to cloud free-riding and pressure to change commercial-use licenses.
FOLLOW THE EVIDENCE